dashboard-edit

name	dashboard-edit
description	Edit an existing chart or its adapter. Opens a session targeting the chart, makes changes, and promotes back to the dashboard.

Edit an existing chart or its adapter.

Before asking anything

Run bin/session-status. If an incomplete session exists, ask to resume or discard (same as dashboard:new Step 1).

Read .session/context.json if present — this gives the active dashboard slug.

If no context.json, determine the target dashboard from:

• The chart-id or slug provided as an argument
• If ambiguous, list dashboards and ask

Step 0 — Determine target dashboard

If the target dashboard slug is not already known, list existing dashboards/*/dashboard.json files, then:

• If the argument identifies a chart-id unambiguously (it exists in exactly one dashboard), use that dashboard.
• Otherwise, ask via AskUserQuestion: "Which dashboard contains the chart you want to edit?"

Write .session/context.json:

{ "dashboard": "<slug>" }

Step 1 — Select chart

If no chart-id was provided as an argument, list all charts from dashboards/<slug>/dashboard.json:

Charts in <slug>:
  1. <title> (id: <id>)
  2. <title> (id: <id>)
  ...

AskUserQuestion: "Which chart do you want to edit?"

Step 2 — Load artifacts into .session/

Silently populate .session/ from the existing chart:

1. Copy dashboards/<slug>/adapters/<adapter-slug>.py → .session/adapter.py
2. Extract the chart config from dashboards/<slug>/dashboard.json["charts"][chartId] → write .session/charts.json (as a single-item array)
3. Find the chart's current placement in dashboards/<slug>/dashboard.json["sections"] → write .session/placement.json
4. Write .session/datasource.json with "is_edit": true and "replaces_chart_id": "<chartId>"

Run the current adapter with the correct env var to get fresh data for context:

DASHBOARD_DIR=dashboards/<slug> python .session/adapter.py

Step 3 — Diagnose and propose

Ask a single open question:

"What would you like to change about ?"

From the answer, immediately determine the scope and do the work silently:

• Data/query change → modify .session/adapter.py, re-run it with DASHBOARD_DIR=dashboards/<slug>, inspect output, regenerate ECharts option if data shape changed
• Visual/chart type change → update the ECharts option in .session/charts.json
• Placement change → update .session/placement.json
• Multiple changes → do all of them

Then open preview (bin/start --preview) which opens /d/<slug>/preview and propose what changed:

Here's what I changed:

  Adapter:  <describe change, or "unchanged">
  Chart:    <describe change, or "unchanged">
  Placement: <describe change, or "unchanged">

Open preview: http://localhost:5173/d/<slug>/preview
Does this look right?

Iterate until confirmed.

Step 4 — Confirm and promote

Show a summary noting this is an update:

Ready to update dashboard:

  Dashboard: <slug>
  Chart:    <title> (replacing existing id: <id>)
  Adapter:  dashboards/<slug>/adapters/<slug>.py (overwritten)
  Section:  <name> (layout unchanged)

Promote?

AskUserQuestion: Yes, promote | Go back and adjust

On approval: run bin/promote. The existing chart entry is replaced.

Guardrails

• Never write to dashboard.json directly.
• Never run bin/promote without confirmation.
• is_edit: true in datasource.json tells bin/promote to replace rather than append.
• Adapters MUST use DASHBOARD_DIR env var for output paths.
• Always write .session/context.json before other artifacts.